Over the past several months, I have been having a lot of conversations. Dozens of large organisations — telcos, cloud communications companies, contact centre platforms, network technology vendors — and a few dozen startups and scaleups operating in and around this space. In the last few weeks alone: Paris, Madrid, New York. London is next week.

This is partly a roadshow — CPaaSAA and our partner Sandbox Industries are actively exploring what a smarter, more collaborative approach to navigating this market looks like, and who wants to be part of it. But the conversations started long before that. We have been listening, testing ideas, reflecting, and puzzling over the same set of questions for the better part of a year.

By now, having spoken with thirty-plus large organisations and a similar number of early-stage companies, a clear pattern has emerged. And this blog is my attempt to articulate it — halfway through the roadshow, while it is fresh.

The pattern is this: everyone knows AI is changing everything. The question — the one nobody has a clean answer to — is: now what?

If you are not OpenAI, not NVIDIA, not one of the handful of companies actually defining the AI era from the centre of it — what is your play? What can you realistically do? What are your actual options?

That is what this is about.

Three Types of Organisation. The Same Problem.

We focus on three categories, because these are the ones we work with and the ones where the disruption is most acute.

The top hundred or so telcos globally. The top twenty cloud communications companies — CPaaS platforms, CCaaS vendors, the companies that own the programmable communications layer. And the larger technology companies — network equipment vendors, enterprise software players, infrastructure providers — operating at the intersection of communications and AI.

Within each of these categories, there is a spectrum — and it matters for understanding who this is actually for.

If you are in the top one or two percent of your category — a Twilio in cloud comms, a Cisco in technology, a tier-one global operator — you probably have the scale, the balance sheet, and the internal capability to navigate this largely on your own. Money is not the binding constraint. Access is not the binding constraint. Orange launched a serious corporate venture operation several years ago and has built real capability. Cisco’s investment and acquisition programme spans hundreds of deals and employs a large dedicated team. These organisations have the firepower to play the game at scale. They may still find value in what CPaaSAA brings — the cross-sector intelligence, the specific focus on the telco-comms-AI intersection — but structurally, they can compete.

That is a very small group. For the next ten to twenty percent — solid, established, meaningful businesses in their categories — the picture is very different. You have real assets and real market position. You do not have unlimited capital or global reach. You cannot go to your board and say “we are doing a hundred-million-dollar acquisition” the way Cisco can. You need a programme, a rationale, short-term evidence that it is working, and board approval at every meaningful step. The constraint is not ambition. It is structure. And the long tail below that? Largely waiting — consciously or not — for the consolidation that is coming.

Our focus, and the focus of this piece, is on that middle tier. The organisations with enough scale to matter and enough at stake to act — but not enough firepower to simply buy their way through the transition. The ones whose boards want to see results this year, not in five.

The dilemma shows up differently depending on where you sit. But the underlying pressure is identical.

The telco. You have structural assets that matter enormously in an AI world — network proximity, regulatory trust, identity infrastructure, real relationships with enterprises and consumers at scale. The question is how you turn those assets into genuine AI differentiation before someone else packages them better than you do. By definition, you have limited geographic coverage. Buying a global AI player rarely makes sense — the fit is wrong, the cost is enormous, and integration is a multi-year project. Building everything internally means competing with companies whose entire existence is software velocity. Putting all your weight behind one obvious platform player — when they are already expensive and the real value has already moved — is exactly what your competitors are doing. That is not differentiation. That is followership. Meanwhile, the trust and identity layer is becoming critical infrastructure for an AI world — silent authentication, fraud prevention, identity-as-a-service. Real commercial opportunities for operators who can credibly own that layer. But which companies are getting it right? How do you know before it is too late and too expensive?

The cloud communications player — mostly CPaaS and contact centre platforms. You are not a passive observer; you are in the middle of it. You see voice AI and agentic automation arriving as a wave of point solutions. You see non-comms players — CRM vendors, enterprise software platforms — entering your space because they need to own the AI that comes out of the customer interaction. Salesforce has a contact centre now. That is not a coincidence. You have an opportunity to evolve your platform from a CPaaS or CCaaS layer into something closer to AI-as-a-service — leveraging your existing channels, your enterprise and SMB relationships, your reach. But which startups do you integrate? Which do you partner with? Which do you get close to early, before they become expensive or inaccessible? Getting this wrong — backing the wrong horse at peak valuation and hoping the roadmap converges — is a strategy that has failed many times before in this industry.

The technology company. In a conversation I had a few days ago with a network technology vendor — a solid, well-established business doing around half a billion in annual revenue — the question was exactly this: the stock has been flat for three years, not because the business is broken, but because the market does not know how to value it in an AI world. They have a software platform. They know AI is coming into their space. They are watching nimbler competitors arrive with lower cost structures. And they are trying to figure out how to get close enough to the startup ecosystem to understand what is actually working — without making a single large bet on something that might be wrong. That conversation is not unique. I have a version of it every week.

Where the Real Action Is

Before getting to what people are trying — and why most of it falls short — it is worth being specific about where I think the value is actually being created. Because understanding what is being built is inseparable from understanding why getting close to it early matters.

AI voice and real-time speech intelligence is not a feature addition. The entire communications layer is being rebuilt around it — from contact centre to enterprise workflow to network provisioning. The companies getting this right are doing something structurally different, not incremental. And the window to back the right ones early is closing.

Agentic AI platforms — the infrastructure for deploying, orchestrating, and governing AI agents across enterprise and network environments — are the new control plane. This is where the real platform battle is being fought. Not at the model layer; models are becoming commodities. The layer above them — orchestration, governance, deployment logic — is where value accrues.

Sovereign AI and regulatory-compliant infrastructure is bigger than most people realise, and it came up repeatedly on this roadshow — including in detailed conversations about what it actually means to deploy AI in a regulated communications environment. GDPR, the EU AI Act, data residency requirements, sector-specific compliance obligations — these make centralised LLM deployment genuinely unworkable for many enterprise use cases. The companies solving this are addressing a fast-growing commercial need, not a compliance checkbox. And given that orchestration layer has to be sovereignty-aware — which model do you run, where does it sit, how do you handle data residency and regulatory jurisdiction — AI orchestration is becoming its own discipline.

Trust, identity, and authentication are being rebuilt for an AI world. Silent authentication, fraud prevention at network level, identity-as-a-service — these are foundational infrastructure for a world where AI agents are acting on behalf of humans at scale. The opportunity is significant. The number of companies genuinely solving it, versus talking about it, is much smaller than the noise suggests.

These are the areas where interesting companies are forming right now. Early stage. Not yet obvious. Not yet expensive. Not yet on the analyst reports. Which is precisely the point.

What People Are Already Trying — And Where It Falls Short

Most organisations are already doing something. Several have tried multiple approaches. Think of it as a spectrum: from doing it entirely yourself at one end, to full acquisition at the other, with a range of collaborative and market-facing approaches in between. Here is what I am observing — and where each model runs into its limits.

Internal innovation programmes. Almost every large organisation has some version of this — an innovation hub, an AI task force, an internal accelerator. The intent is right. The execution almost always runs into the same wall: internal teams are not wired for startup speed. The procurement cycles, the risk appetite, the incentive structures — they are built for a different kind of work. We have direct experience with this model at The Next Cloud, having worked with telco innovation hubs. Even when these programmes run at two or three times the normal pace for a large organisation, they are still moving far too slowly relative to the market. What worked a decade ago is not fit for purpose today.

Telco innovation JVs. Some organisations have tried to address this through joint ventures — structures designed to pool resources across multiple players for shared innovation goals. The instinct is sound. The governance reality is consistently painful: multiple large organisations with competing priorities, board-level complexity, slow decision-making by construction. For certain narrowly-defined problems, this can work. As a model for navigating fast-moving AI disruption, it tends to struggle — particularly when the JV was conceived before AI started reshaping the landscape it was meant to address.

Generic accelerator programmes. I have spent significant time with programmes like Startupbootcamp — rated the number one startup accelerator in Europe for several years, active across more than a dozen countries, with genuine success stories including SendCloud and many others. Y Combinator, similarly, has an extraordinary track record. These programmes work. They are just not built for our niche. Y Combinator is outstanding if you are building a consumer AI product or a horizontal SaaS play. If you are building at the intersection of telco, CPaaS, and AI — with all the complexity that entails around regulation, operator relationships, and enterprise sales cycles — the fit is poor. The programme lacks the sector-specific knowledge, the right network, and the commercial pathways that matter for companies in this space. We explored the possibility of building a telco-focused accelerator at one point. The model would have required multiple telcos as shared investors, global coverage, and a mix of communications and AI players. In other words — it would have needed to become something structurally quite close to what we are building now.

Corporate venture capital. A number of larger telcos have built in-house CVC arms, and many of the people running them are doing genuinely thoughtful work. But there are structural limitations worth naming. A corporate VC attached to a single operator is shaped by that operator’s geography, strategy, and blind spots. It tends to see what comes to it rather than actively mapping the full landscape. And it often lacks the cross-sector perspective that matters most right now — the intersection of telco infrastructure, cloud communications, and AI is where the most interesting things are forming, and that intersection is hard to see clearly from inside one organisation.

Peer groups and informal co-investment clusters. This is the most interesting of the existing approaches, and the most underrated. Groups of European telcos meet regularly to exchange views, share experiences, and collectively look at the market. The instinct is exactly right: pooling intelligence and perspective across peers is more powerful than going it alone. But these groups tend to be relatively closed, geographically concentrated, and still largely telco-centric. When I spoke with one of the major European operators in Paris recently, the honest reflection was: we have some areas we are looking at, but we do not see everything. We are constrained. When you are primarily a telco looking at a market being reshaped by the convergence of cloud communications, AI infrastructure, and network intelligence, a telco-centric lens only gets you so far.

Developer funnels, free credits, and hackathons. This one is specific to CPaaS and cloud communications players, and worth addressing directly because it is so widely practiced. The model is familiar: give startups free platform access, API credits, or developer accounts; run a hackathon; build a startup programme. The theory is that developers who build on your platform will scale on your platform. The execution has a fundamental flaw.

Retention. When a startup that built on your platform during its early, capital-light phase starts to scale and generate real revenue, they evaluate infrastructure on cost, performance, and commercial fit — not on loyalty to the platform that gave them free tokens two years ago.

And the hackathon itself? Let us be honest about what it means in 2025. AI can generate a convincing demo, a working proof of concept, and a polished pitch in hours. Three teams winning prizes for cool demos does not move the needle. What you need is not proof of concept — that is the easy part now. What you need is real solutions generating real revenue with real customers. A hackathon does not get you there. It gets you a prize ceremony, a LinkedIn post, and nine months of your team’s time. No skin in the game. No selection rigour. No sustained engagement through the hard part of building a business.

M&A. Acquire the winners. It is a clean narrative and it works — if you are big enough and you get the timing right. For organisations with global enterprise reach and the balance sheet to match, buying at scale is a viable strategy. For most of the companies I speak with, it is not. And even for the large players, M&A in innovation is harder than it looks.

The communications and AI space has produced a series of instructive examples in recent years. Ericsson spent $6.2 billion to acquire Vonage — one of the largest bets in the CPaaS world — and has since taken impairment charges of around four billion dollars against it, while still working through how to integrate, regionally adapt, and extract strategic value. Mavenir acquired Telestax, one of the most respected open-source communications platforms, and effectively wound it down within a year as strategies shifted. NICE acquired an AI firm for close to a billion and is now in the integration phase. Microsoft acquired Metaswitch, then sold it a few years later to Alianza — whom I spoke with just last week — after losing significant staff and customer confidence. Nokia acquired Rapid (formerly RapidAPI) in November 2024 — a once-billion-dollar-valued API marketplace platform that had seen significant decline — folding it into its Network as Code strategy. The price was not disclosed. Whether the integration delivers the intended value remains to be seen.

And I remember clearly when BT acquired Ribbit in 2008 for $105 million — one of the earliest programmable voice API platforms, a genuinely ahead-of-its-time company. BT acquired it, tried to fold it into a telco operating model, and in the process largely lost what made it interesting. I know one of the leaders from that era well — he stayed with BT for nearly two decades, recently left to found a new startup, and we are now in conversation about that company. The story came full circle. But the lesson from Ribbit did not change: acquiring innovation is not the same as sustaining it.

The pattern is consistent. Studies put M&A failure rates at around eighty percent, and in technology and innovation the reasons are clear: you acquire the company but lose the culture, the speed, and the founding energy that made it worth acquiring. M&A works for buying revenue and market position. It is a much weaker instrument for buying innovation.

What Smart Capital Is Learning

It is worth noting that the venture capital world itself is evolving in response to exactly this dynamic — and the direction of travel is instructive.

The traditional model — identify promising companies, write cheques, wait for exits — is giving way to something far more active. Andreessen Horowitz (a16z), now the largest venture capital firm in the world by assets under management, pioneered what they call the platform model: a large team of operators — in marketing, talent, go-to-market, legal, and regulatory affairs — dedicated entirely to helping portfolio companies grow. Not just capital and a board seat. Active, sustained, expert support across every dimension of company-building.

The results speak for themselves. The firms doing this well are more selective at entry, more engaged throughout, and generate better outcomes for founders and investors alike. The model has shifted the entire industry’s expectations of what a capital partner should do.

This matters for our conversation. If even generalist VC is moving toward active value-add, then a sector-specific vehicle — one where the investors are also the potential customers, partners, and distribution channels for the companies being backed — has a structural advantage that is very hard to replicate. The startups get more than capital. They get market access, go-to-market support, and relationships inside an industry that is notoriously difficult to break into from the outside. The investors get more than a financial stake. They get first-mover intelligence, early commercial relationships, and a seat at the table during the formation period.

If You Are Not One of the Twenty Largest Companies in the World

Let us be direct about something.

The Ciscos, the NVIDIAs, the hyperscalers — they can spend hundreds of millions on acquisitions without blinking. Global enterprise focus, massive balance sheets, armies of corporate development people. For them, buying the winner at any price is a viable strategy.

For most of the organisations I talk to, that is not the game. And trying to play it is how you lose.

If you do not have that kind of wallet, you need to be smarter. You need to team up. You need to work with peers facing the same pressures and navigating the same questions. You need access to early-stage companies — in AI voice, agentic platforms, sovereign infrastructure, trust and identity — before they are on the analyst reports, before the valuations reflect the hype, before everyone else has made their move.

This is not a long-term play dressed up as short-term strategy. The intelligence you get from being close to what is forming has immediate commercial value. You are meeting founders. You are seeing technology before it is public. You are building relationships with the companies solving problems on your roadmap — and influencing that roadmap, not just reacting to it. That is a quarterly win. That is a board story for today.

Nobody in this industry is thinking about a five or seven year horizon right now. The pace of change, the volume of capital flowing into AI, the speed at which competitive positions are being established and disrupted — this is a market unlike anything we have seen. Short-term wins are not a preference. They are a requirement.

And the FOMO dimension is real. If you are not actively and intelligently close to the early-stage companies shaping this space — if you are not at the table — someone else is. Your competitors are getting the market intelligence. They are building the early relationships. They are seeing opportunities before they are public and making informed moves while you are still discussing whether to commission another report. The ecosystem advantage compounds. The longer you wait to be part of it, the more expensive and difficult it becomes to catch up.

So What Would the Right Structure Actually Look Like?

I want to end with a question rather than a sales pitch — because I think the question is more useful right now.

If you accept the premise that doing it alone is too slow, that generic accelerators lack the sector depth, that CVC misses the cross-sector picture, that hackathons are theatre, and that M&A is out of reach for most and unreliable even for those who can afford it — what would the right structure actually look like?

My working answer, built from months of these conversations: it would need to be collective. A group of organisations — telcos, cloud communications players, technology companies — pooling resources to build the intelligence engine none of them can afford to build alone. Scouting what is forming. Understanding which startups are solving real problems versus generating impressive demos. Feeding that intelligence back to the participants in a way that is immediately actionable.

It would need to go beyond intelligence. It would need to enable meaningful participation in the companies that look most promising — early, when the stakes are still reasonable, when the founders still need strategic partners and not just capital. And it would need to bring something the founders actually want: not just money, but market access. Go-to-market support. Enterprise relationships inside an industry that is notoriously hard to enter from the outside.

And it would need to be financially structured so that this is not an expense but an investment — where collective participation generates real returns, spread across a portfolio of bets rather than concentrated in a single high-risk call.

The telcos, cloud communications platforms, and technology companies I have been speaking with have all the ingredients: market relationships, go-to-market reach, deep domain knowledge, and a genuine stake in which companies succeed. What most of them lack is the structure to deploy those assets collectively and intelligently.

Building that structure — creating the engine that connects incumbents with the startups shaping the next decade of this industry — is what CPaaSAA exists to do. And it is exactly what we are working on.

More from London next week.

Rob Kurver is Founding Partner of CPaaSAA and Founder & CEO of The Next Cloud — a strategy practice helping organisations navigate AI, cloud communications, and the next generation of digital infrastructure.